1. This comment applies to malicious software ("malware") that's installed unwittingly by the victim of a network attack. It does not apply to software, such as keystroke loggers, that may be installed deliberately by an intruder who has hands-on access to the victim's computer. That threat is in a different category, and there's no easy way to defend against it. If you have reason to suspect that you're the target of such an attack, you need expert help.
2. All versions of OS X since 10.6.7 have been able to detect known Mac malware in downloaded files, and to block insecure web plugins. This feature is transparent to the user, but internally Apple calls it "XProtect." The malware recognition database is automatically checked for updates once a day; however, you shouldn't rely on it, because the attackers are always at least a day ahead of the defenders.
The following caveats apply to XProtect:
It can be bypassed by some third-party networking software, such as BitTorrent clients and Java applets.
It only applies to software downloaded from the network. Software installed from a CD or other media is not checked.
3. Starting with OS X 10.7.5, there has been a second layer of built-in malware protection, designated "Gatekeeper" by Apple. By default, applications and Installer packages downloaded from the network will only run if they're digitally signed by a developer with a certificate issued by Apple. Software certified in this way hasn't actually been tested by Apple (unless it comes from the Mac App Store), but you can be reasonably sure that it hasn't been modified by anyone other than the developer. His identity is known to Apple, so he could be held legally responsible if he distributed malware. For most practical purposes, applications recognized by Gatekeeper as signed can be considered safe.
Gatekeeper doesn't depend on a database of known malware. It has, however, the same limitations as XProtect, and in addition the following:
It can easily be disabled or overridden by the user.
A malware attacker could get control of a code-signing certificate under false pretenses, or could find some other way to evade Apple's controls.
4. Starting with OS X 10.8.3, a third layer of protection has been added: a "Malware Removal Tool" (MRT). MRT runs automatically in the background when you update the OS. It checks for, and removes, malware that may have evaded the other protections via a Java exploit (see below.) MRT also runs when you install or update the Apple-supplied Java runtime (but not the Oracle runtime.) Like XProtect, MRT is presumably effective against known attacks, but maybe not against unknown attacks. It notifies you if it finds malware, but otherwise there's no user interface to MRT.
5. Beyond XProtect, Gatekeeper, and MRT, there’s no evidence of any benefit from other automated protection against malware. The first and best line of defense is always your own intelligence. With the possible exception of Java exploits, all known malware circulating on the Internet that affects a fully-updated installation of OS X 10.6 or later takes the form of so-called "trojan horses," which can only have an effect if the victim is duped into running them. The threat therefore amounts to a battle of wits between you and the malware attacker. If you're smarter than he thinks you are, you'll win.
That means, in practice, that you never use software that comes from an untrustworthy source. How do you know whether a source is trustworthy?
Any website that prompts you to install a “codec,” “plug-in,” "player," "extractor," or “certificate” that comes from that same site, or an unknown one, is untrustworthy.
A web operator who tells you that you have a “virus,” or that anything else is wrong with your computer, or that you have won a prize in a contest you never entered, is trying to commit a crime with you as the victim. (Some reputable websites did legitimately warn visitors who were infected with the "DNSChanger" malware. That exception to this rule no longer applies.)
Pirated copies or "cracks" of commercial software, no matter where they come from, are unsafe.
Software of any kind downloaded from a BitTorrent or from a Usenet binary newsgroup is unsafe.
Software with a corporate brand, such as Adobe Flash Player, must be downloaded directly from the developer’s website. If it comes from any other source, it's unsafe.
Regardless of version, experience has shown that Java on the Web can't be trusted. If you must use a Java applet for a specific task, enable Java only when needed for the task and disable it immediately when done. Close all other browser windows and tabs, and don't visit any other sites while Java is active. Never enable Java on a public web page that carries third-party advertising. Use it, when necessary, only on well-known, login-protected, secure websites without ads. In Safari 6 or later, you'll see a lock icon in the address bar with the abbreviation "https" when visiting a secure site.
Follow the above guidelines, and you’ll be as safe from malware as you can practically be. The rest of this comment concerns what you should not do to protect yourself from malware.
7. Never install any commercial "anti-virus" or "Internet security" products for the Mac, as they all do more harm than good, if they do any good at all. If you need to be able to detect Windows malware in your files, use the free software ClamXav — nothing else.
Why shouldn't you use commercial "anti-virus" products?
Their design is predicated on the nonexistent threat that malware may be injected at any time, anywhere in the file system. Malware is downloaded from the network; it doesn't materialize from nowhere.
In order to meet that nonexistent threat, the software modifies or duplicates low-level functions of the operating system, which is a waste of resources and a common cause of instability, bugs, and poor performance.
By modifying the operating system, the software itself may create weaknesses that could be exploited by malware attackers.
8. ClamXav doesn't have these drawbacks. That doesn't mean it's entirely safe. It may report email messages that have "phishing" links in the body, or Windows malware in attachments, as infected files, and offer to delete or move them. Doing so will corrupt the Mail database. The messages should be deleted from within the Mail application.
ClamXav is not needed, and should not be relied upon, for protection against OS X malware. It's useful only for detecting Windows malware. Windows malware can't harm you directly (unless, of course, you use Windows.) Just don't pass it on to anyone else.
A Windows malware attachment in email is usually easy to recognize. The file name will often be targeted at people who aren't very bright; for example:
ClamXav may be able to tell you which particular virus or trojan it is, but do you care? In practice, there's seldom a reason to use ClamXav unless a network administrator requires you to run an anti-virus application.
9. The greatest harm done by security software, in my opinion, is in its effect on human behavior. It does little or nothing to protect people from emerging threats, but they get a false sense of security from it, and then they may behave in ways that expose them to higher risk. Nothing can lessen the need for safe computing practices.
10. It seems to be a common belief that the built-in Application Firewall acts as a barrier to infection, or prevents malware from functioning. It does neither. It blocks inbound connections to certain network services you're running, such as file sharing. It's disabled by default and you should leave it that way if you're behind a router on a private home or office network. Activate it only when you're on an untrusted network, for instance a public Wi-Fi hotspot, where you don't want to provide services. Disable any services you don't use in the Sharing preference pane. All are disabled by default.